Figure 2: Overview of the proposed scene parsing framework.